SBIR Phase I: Rapid and Accurate Multi-Variable Optimization Software for Arrays of Heat Sinks

这个小企业创新研究（SBIR）项目的更广泛的影响/商业潜力是减少冷却驱动互联网时代的数据和电信中心所需的电力。目前，美国近2.0%的能源消耗用于运行此类数据中心，其中一半通常用于冷却。相应的温室气体排放也会减少。从长远来看，这项技术可以用于提高空调/制冷和发电的效率。商业上，该产品将参与的计算流体动力学软件市场预计将从目前的13亿美元增长到2022年的20多亿美元。它还将刺激具有最佳几何形状的高效热管理的特种散热器制造的增长。这个SBIR一期项目的智力优势是基于提出了三种常见数值方法的混合，即计算流体动力学（CFD）、流动网络建模（FNM）和多变量优化（MVO），这是一个重要的目标。结果将是一个软件平台，可以准确地同时优化各种类型电路包（例如数据中心的刀片服务器）中散热片阵列的几何形状。单独使用CFD是不可能的，因为它太慢了；单独使用FNM是不可能的，因为它不够准确。保持CFD精度的混合方法是在FNM中嵌入CFD模拟中预先计算的、无量纲的流阻和热阻查找表。该软件将容纳空气和水的单相流动以及水和制冷剂的蒸发/沸腾流动，这需要模拟复杂的热量，质量和动量传输现象。实验将验证其在电路封装热模型上的准确性。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) project is to reduce the power required to cool the data and telecommunications centers that drive the internet age. Currently nearly 2.0% of the energy consumed in the United States is used to run such data centers and, often, half of this is used for cooling. The corresponding greenhouse gas emissions will also be reduced. Longer term this technology can be adapted to increasing the efficiency of air-conditioning/refrigeration and power generation. Commercially, the computational fluid dynamics software market the product will compete in is projected to grow from $1.3 billion at present to over $2 billion by 2022. It will also stimulate the growth of the manufacture of specialty heat sinks with optimal geometries for efficient thermal management.This intellectual merit of this SBIR Phase I project is based on proposing to hybridize three common numerical methods, i.e., computational fluid dynamics (CFD), flow network modeling (FNM) and multi-variable optimization (MVO), a non-trivial objective. The result will be a software platform that can accurately and simultaneously optimize the geometry of an array of heat sinks found in various types of circuit packs, e.g., blade servers in data centers. This is not possible with CFD alone as it is too slow and it is not possible with FNM alone as it is not sufficiently accurate. The method of the hybrid approach to preserve the accuracy of CFD is to embed pre-computed, non-dimensional look-up tables of flow and thermal resistances from CFD simulations in an FNM. The software will accommodate both single-phase flows of air and water and evaporating/boiling flows of water and refrigerant, which requires modeling complex heat, mass and momentum transport phenomena. Experiments will validate its accuracy on thermal mock-ups of circuit packs.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.